This code should run provided you are able to load the datasets in correctly. Otherwise, view the HTML.
mutate makes new columns in a data frame based on existing columns
Answer these questions with a single chain of commands:
First, we must read in the packages required for this document.
library(tidyverse)
library(readxl)Read in the TCGA dataset, the exact location depends on you. If you save the Rmd and the csv in the same location, you may load it purely by name as is done below.
tcga <- read_csv("tcga_stats.csv")
tcgaIt is possible to answer these questions in a single chain of commands. This takes the tcga dataset, makes two new columns
bio.total is the sum of copy_number_variation and dna_methylation
new.gender is TRUE if “male” or “female” is found in the gender column, and FALSE if anything else
Then we filter based on the new columns, reducing the dataset to only entries where gender is “male” or “female” and bio.total is greater than 10.
# take the tcga dataset
tcga %>%
# make two new columns, bio.total and new.gender
mutate(bio.total = copy_number_variation + dna_methylation,
new.gender = ifelse(gender == "female" | gender == "male", TRUE, FALSE)) %>%
# filter the new data frame to rows that match these parameters
filter(new.gender == TRUE & bio.total > 10)summarise allows you to create custom summary functions on grouped datasets. summarise creates new columns and gets rid of ones you’re not using. It takes the argument of a dataset, the name of a new column and the way that column is defined. It always results in a data frame. This is an example, it won’t run.
df %>%
group_by(some_variable) %>%
summarise(new_col_name = way_to_get_new_col) %>%
ungroup()For example, if we want to know the mean number of transcriptome files for each for each primary cancer site, we can do the following
# Take the tcga dataset
tcga %>%
# group by primary cancer site
group_by(primary) %>%
# define a custom function which takes the mean of the transcriptome column for each group
summarise(mean.trans.files.per.primary = mean(transcriptome_profiling, na.rm = TRUE)) %>%
ungroup()Answer these questions with one chain of commands:
# take the tcga dataset
tcga %>%
# group by primary cancer site
group_by(primary) %>%
# create two summary columns, the sum of all clincal files per group and
# the sum of all biospecimen files per group
summarise(clin.sum = sum(clinical, na.rm = TRUE),
bio.sum = sum(biospecimen, na.rm = TRUE)) %>%
ungroup() %>%
# add a new column that calculates the ratio of the two
mutate(ratio = clin.sum / bio.sum)This method does the same thing but in a more condensed manner.
tcga %>%
group_by(primary) %>%
summarise(ratio = sum(clinical, na.rm = TRUE) / sum(biospecimen, na.rm = TRUE)) %>%
ungroup()dir simply tells you the contents of a directory. The only necessary argument is a path to a directory in quotes. If none is supplied, it thinks you’re asking about the directory where you’re running your code from (where the Rmd is saved). The results of running this will vary depending on where you save your Rmd.
dir()To find files with particular extensions, you can supply a pattern argument. To find all .xlsx files
dir("path/to/directory", pattern = ".xlsx")To get the full file path, include full.names = TRUE, this is necessary if you plan to load files using the information derived from dir. This would get you all of the full filepaths to all xlsx files in this made up directory. You could use these as input to read_xlsx.
dir("path/to/directory", pattern = ".xlsx", full.names = TRUE)recursive = TRUE will tell dir to look in subdirectories for files too.
This function simply takes the path to an excel file and returns the sheet names as a vector.
# get the path to any excel files in this directory, there's only 1
my_excel_file <- dir(".", pattern = ".xlsx", full.names = TRUE)
# pass the path excel_sheets to figure out the sheet names
my_excel_file %>%
excel_sheets()## [1] "A+1" "A+2" "A+3" "A+4" "A+5" "A+6" "A-1"
## [8] "A-2" "A-3" "A-4" "A-5" "A-6" "REL606" "REL607"set_names assigns names to an object, or elements of an object. If no names are supplied, it names a vector with itself. Each element of the sheet names vector is named with itself. The elements of the vector are in quotes and the names are not.
my_excel_file %>%
excel_sheets() %>%
set_names()## A+1 A+2 A+3 A+4 A+5 A+6 A-1 A-2
## "A+1" "A+2" "A+3" "A+4" "A+5" "A+6" "A-1" "A-2"
## A-3 A-4 A-5 A-6 REL606 REL607
## "A-3" "A-4" "A-5" "A-6" "REL606" "REL607"This function takes the path to an excel file and reads it in as a data frame. It can choose a specific sheet if you ask for it, but otherwise reads in the first sheet with little indication of that. Remember, the skip, na, and col_names arguments are specific to this excel file and might not work for one you have.
read_xlsx(path = my_excel_file, # the excel file is here
col_names = TRUE, # the first row after the skip is column names
na = "NaN", # NAs are written as NaN in this data
skip = 1) # skip the first rowWe just learned about dir(), excel_sheets(), and read_xlsx(), use them to:
Simply use dir to find the file path of the .xlsx file.
excel_file_data <- dir(".", pattern = ".xlsx", full.names = TRUE)
excel_file_data## [1] "./cell_size1.xlsx"Pass the file path to excel_sheets, either of these ways is fine. set_names is unnecessary here.
excel_file_data %>%
excel_sheets()## [1] "A+1" "A+2" "A+3" "A+4" "A+5" "A+6" "A-1"
## [8] "A-2" "A-3" "A-4" "A-5" "A-6" "REL606" "REL607"excel_sheets(excel_file_data)## [1] "A+1" "A+2" "A+3" "A+4" "A+5" "A+6" "A-1"
## [8] "A-2" "A-3" "A-4" "A-5" "A-6" "REL606" "REL607"Pass our excel file path to the read_xlsx function with the given parameters.
read_xlsx(path = excel_file_data, skip = 1, na = "NaN", col_names = TRUE)You can do this all in one chain of commands
# take the file path
excel_file_data %>%
# determine the sheet names, returns a vector
excel_sheets() %>%
# take the 6th element of that vector
.[6] %>%
# pass that sheet name to read_xlsx, which automatically knows it's a sheet name
read_xlsx(path = excel_file_data, skip = 1, na = "NaN", col_names = TRUE)Reading in multiple sheets requires the use of the map function. map applies a function to elements of a vector and results in a list, one item generated for each element of the vector. In this case, we read each of the 14 sheets in as their own data frame and end up with a list of 14 data frames. Importantly, we do want the vector of sheet names to have names, so we use set_names.
# find our file
file.loc <- dir(".", pattern = ".xlsx")
file.loc## [1] "cell_size1.xlsx"# take out file path
df.list <- file.loc %>%
# get the sheet names
excel_sheets() %>%
# name the vector using itself
set_names() %>%
# for each item in that vector, read in that sheet from our
# excel file as a data frame and store it in a list
map(read_xlsx, path = file.loc, col_names = TRUE, skip = 1, na = "NaN")
# Uncomment this if you want to see the list of data frames, 14 dfs renders poorly in HTML
#df.listImportantly, note that the items in the list all have the same name as the sheet they came from. This is not the case if you omit set_names
df.list.sans.names <- file.loc %>%
# get the sheet names
excel_sheets() %>%
# for each item in that vector, read in that sheet from our
# excel file as a data frame and store it in a list
map(read_xlsx, path = file.loc, col_names = TRUE, skip = 1, na = "NaN")
# commented out for the same reason, 14 dfs
#df.list.sans.namesbind_rows stacks data frames on top of each other and only works if the data frames have the same number and names of columns. If fed a list, it knows you want to stack all the data frames in the list together. the .id argument allows you to use the names of data frames in a list in a new column so you know which rows came from where.
Simply bind_rows the list
bind_rows(df.list)Now, you must use the .id argument. Note the addition of the sample column.
all.data <- bind_rows(df.list, .id = "sample")
all.dataNow that we have a single data frame, this is easy. group by sample and use summarise to make a mean SHAPE.angularity column.
all.data %>%
group_by(sample) %>%
summarise(mean_shape_angularity = mean(SHAPE.angularity, na.rm = TRUE)) %>%
ungroup()Using our single data frame, we can mutate in a new column that back calculates radius from area of a perfect circle. This new column appears at the right most end of the data frame.
all.data %>%
mutate(radius = sqrt(SHAPE.area/pi))We can also join data frames side by side. left_join and right_join perfomr this action. They are similar, but different.
Given these two dataset, which have Gene in common between, we would like a single data frame that represents all the data.
genes1 <- read_csv("genes1.csv")## Parsed with column specification:
## cols(
## Gene = col_character(),
## Length = col_double(),
## RPF = col_double(),
## mRNA = col_double(),
## TE = col_double()
## )genes2 <- read_csv("genes2.csv")## Parsed with column specification:
## cols(
## Gene = col_character(),
## FEcap = col_double(),
## uATG = col_logical(),
## utr = col_double(),
## gc = col_double()
## )genes1genes2For these data frames which each contain all of the genes, left and right joins are the same. Though the exact positions of the rows differ between the methods (this could be overidden by arranging each result by the same thing), all of the data is in both of them.
left_join(genes1, genes2, by = "Gene")right_join(genes1, genes2, by = "Gene")If one of the datasets is missing some of the genes, for example, we have a new dataset that is only long genes
long.genes <- genes1 %>%
filter(Length > 900)
long.genesAnbd we still want to join the rest of the information in genes2 to it, left_join and right_join then produce different results.
This attempts to join, where possible, all matching rows from genes2 to long.genes. The join is to the left in that regard because it’s genes2 to long.genes. There’s only 562 matching rows, so those are the only ones it adds
left_join(long.genes, genes2, by = "Gene")right_join has the opposite effect. This tries to join all rows from long.genes to genes2. Where it find mathcing rows, we get a complete set of data. Where there is no match, is jsut puts NA in that cell.
right_join(long.genes, genes2, by = "Gene")These results we then be reversed if we reversed the order in which we fed the data frames to the function. Notice that the row counts are different if we reverse the order.
# originally gave us 562 rows
left_join(genes2, long.genes, by = "Gene") %>%
nrow()## [1] 4839# originally gave us 4839 rows
right_join(genes2, long.genes, by = "Gene") %>%
nrow()## [1] 562